Dangshan pear is one of the most popular fruits in China. Among them, the surface defects can often be required to monitor the pear in real time. However, the practical requirements are limited to the computing resources of the devices under real-time constraints in industrial environments. In this study, a lightweight and efficient detection model (named AIC-YOLOv11n) was developed using the YOLOv11n architecture. Specifically, an Adown down-sampling module was introduced into the backbone. Both the floating-point and parameters were reduced to enhance the feature extraction. Additionally, the original C2PSA module was replaced with the C2PSA-iRMB one. An inverted residual mobile block (iRMB) was integrated with the attention mechanisms in order to efficiently capture the long-range dependencies with less computational overhead. Moreover, a cross-scale feature fusion module (CFFM) was employed in the neck structure of the network. Some features at different scales were effectively merged to improve the detection accuracy of the small-scale defects. A dataset with 5,000 labeled images was constructed to validate the performance of the improved model. The images were also collected using the conveyor-belt multi-surface imaging system, that equipped with synchronized upper and lower illumination boxes and industrial-grade cameras. The dataset included five categories: Calyx, stem-end cap, scratches, rust spots, and mold spots. Data augmentation was also carried out, including rotation, flipping, and brightness adjustments. The dataset was then partitioned into the training, validation, and test datasets at an 8:1:1 ratio. Experimental results showed that the improved AIC-YOLOv11n model achieved better performance in detection, compared with the baseline YOLOv11n. Specifically, there was a precision of 92.5%, a recall rate of 87.5%, an mAP0.5, of 92.7%, and an mAP0.5-0.95 of 70.5%, which were improved by 0.3, 5.5, 5.1, and 2.4 percentage points, respectively. Additionally, the computational costs were reduced significantly to require only 4.3 G, 1.46 million parameters, and a model size of 3.11 MB, which were reduced by 31.7%, 43.4%, and 40.5%, respectively. Furthermore, the peak GPU memory usage remained below 4.83 GB, and the inference speed reached 120.1 frames per second (FPS), thus fully meeting the real-time requirement of the defect inspection. Ablation studies demonstrated that there were the great contributions of the three modules. Among them, the Adown achieved the greatest improvement in the recall, while the CFFM significantly enhanced the detection accuracy of the small objects, and C2PSA-iRMB effectively increased the precision. Grad-CAM visualization further confirmed that the improved model was focused accurately on the defect regions, while suppressing the interference from normal anatomical structures. Online TensorRT deployment was then utilized to validate the improved model in an industrial scenario. Once converted to a TensorRT FP16 inference engine, there was a single-image inference latency of just 1.4 ms without compromising accuracy, indicating its suitability for real-world applications. In conclusion, the AIC-YOLOv11n was provided to balance the accurate, efficient, and lightweight surface defect detection on Dangshan pears. Model pruning, knowledge distillation, and transfer learning can be expected for the more fruit types in agricultural industries.
- Article type
- Year
- Co-author
Smart agriculture is a promising potential direction in modern agriculture worldwide. However, China’s agriculture is still limited by multiple constraints at present, such as resources, markets, and the environment. Therefore, smart agriculture can be expected to take as a breakthrough point, in order to promote the agricultural transformation and modernization. In this study, a systematic analysis was made on the current status of the research and application of the key technologies and equipment in smart agriculture. The main challenges were clarified to prospect the development trend and promotion path of China’s smart agriculture in the future. Specifically, 1) Smart agriculture serves as the modern advanced form to take the information and knowledge as the core elements. Smart agriculture aimed to realize the digital management, intelligent decision-making, automatic operation, precise input, and networked services, after the deep integration of advanced productive forces, such as modern information, industrial equipment, and agricultural biotechnology. 2) China’s smart agriculture relied mainly on three common supports: sensor and perception systems, decision-making models and algorithms, as well as intelligent platforms. Remarkable progress was made in the R＆D of sensor technologies, and the agricultural satellite remote sensing, as well as the agricultural unmanned aerial vehicle remote sensing. Agricultural basic algorithms were continuously optimized to couple the animal and plant models, gradually shifting towards a multi-factor system. The R＆D of large agricultural artificial intelligence models was to facilitate the intelligent platform architecture. Data storage and management technologies were developed to integrate the application using knowledge graphs, indicating the hot spot. 3) The application scenarios of smart agriculture in China were constantly expanding under advanced technology. a) The frontier technologies, such as artificial intelligence and gene editing, were promoted to transform the breeding paradigm towards intelligence. b) The technical system of smart farms was gradually improved to initially achieve ‘machine replacing people’. c) Smart facility agriculture has advanced significantly to produce the core equipment in the construction of a high-end intelligent greenhouse. d) The livestock and poultry breeding have achieved significant advances in smart technology. The smart farms for dairy, pork, and poultry were at the top worldwide. e) Smart fishery was limited to the sensors, growth models, and special robots. A number of smart fish farms were constructed to accelerate industrialization. f) Smart agricultural product supply chains were still in the initial stage. The partial breakthroughs and application demonstrations were also observed in post-harvest processing, quality traceability, and smart storage and transportation. 4) There were still many challenges in smart agriculture, in terms of the key technology R＆D, technology transfer, and industrialization, and basic support systems. a) The basic model algorithms were insufficient to substitute for the high-end sensors and robots. b) The scientific and technological achievements were necessary to translate into practical applications. The chain ecosystem of the digital industry was also required to evaluate the integration and sustainability of application scenarios. c) The decision-making on the data resource was also required to strengthen the professional talents and farmers’ digital literacy. 5) Looking to the future, five strategic directions were recommended in China’s smart agriculture: the low-cost and high-precision agricultural information perception, agricultural artificial intelligence, agricultural low-altitude economy, small-scale smart agriculture, and ‘zero-carbon’ smart agriculture. The institutional framework and ‘digital foundation’ of smart agriculture should be consolidated to promote smart agriculture in China. Universities and research institutions should promote the R＆D of key technologies for smart agriculture and the cultivation of high-end talents. Technology-based enterprises should construct a market-oriented and sustainable smart agriculture industrial cluster. Agricultural business entities and farmers also need to promote the practical application of ‘scenario + chain’ in smart agriculture.
Swarm intelligence in agricultural robotics has emerged as a strategic frontier that can simultaneously boost productivity and sustainability in modern agriculture. By enabling safe, reliable, autonomous, and efficient operations across open, dynamic, and partially structured farm environments, swarm intelligence provides a unifying paradigm for transitioning from single-robot autonomy to coordinated multi-robot systems. This paper presents a systematic review of the fundamental connotation, elements, key technologies, and future directions of agricultural robot swarm intelligence. We first clarify the concept and origins of swarm intelligence and summarize its defining properties—distributed control, self-organization, and dynamic adaptability—then propose a domain-specific conceptual framework for agricultural robot swarms centered on “Individual autonomy–Information sharing–Swarm collaboration.” We then delve into the four critical technological pillars underpinning agricultural robot swarm intelligent (ARSI): collaborative perception, collaborative planning, collaborative control, and ground-air cross-domain collaboration. For collaborative perception, we discuss multi-robot localization and mapping in non-structured fields, emphasizing robustness under GNSS multipath, vegetation occlusion, and seasonal appearance changes. Representative SLAM and semantic mapping strategies are compared, including centralized and fully distributed map fusion, bandwidth-aware loop closure, and spatiotemporal synchronization across heterogeneous sensors. We highlight advances that fuse LiDAR-vision-inertial data, improve robustness in highly dynamic scenes, and maintain long-term map consistency for seasonal operations. In collaborative planning, we review task allocation and motion planning under large task scales, heterogeneous robot capabilities, and tight operation windows. Four families of task assignment methods are contrasted—exact optimization, market/auction mechanisms, bio-inspired metaheuristics, and learning-driven approaches—together with their trade-offs in optimal, scalability, and responsiveness. For coverage and path planning, we summarize graph-search, sampling-based, potential-field, and intelligent optimization methods, and discuss trajectory tracking for synchronized multi-robot execution using techniques such as artificial potential fields, spline smoothing, and model predictive control. For collaborative control, we briefly review formation and consensus methods (leader-follower, behavior-based control, event-triggered schemes, and predictive control) that ensure safety, precision, and resource efficiency in long-duration field operations. Special attention is given to region-reaching consensus for multi-task area operations, event-triggered formation control that reduces computation/communication loads while preserving tracking accuracy, and robust schemes that accommodate model uncertainties. We then analyze ground–air cross-domain collaboration, wherein UAVs provide wide-area monitoring, semantic mapping, and communication relays, while UGVs execute high-precision, energy-efficient interventions. We discuss registration and fusion between aerial and ground maps, closed-loop pipelines from perception to targeted actuation (e.g., site-specific weeding/spraying), and the scalability challenges of communication, safety, and real-time scheduling in large deployments. In addition, we present compelling case studies from global initiatives and commercial deployments, such as the RHEA project’s weed control system, SwarmFarm’s cloud-optimized fleet, and Carbon Robotics’ laser-weeding technology, demonstrating tangible benefits in reducing chemical inputs and labor costs. Finally, we outline five key future research directions: 1) the development of domain-specific agricultural foundation models to drive knowledge-informed decision-making; 2) the creation of comprehensive “Farm-Digital Twin” platforms for closed-loop simulation and re-planning; 3) the adoption of hierarchical edge-cloud architectures to manage computational and communication bottlenecks; 4) the advancement of dynamic, game-theoretic planning frameworks for highly uncertain environments; and 5) the standardization and large-scale validation of ground-air collaborative systems. This synthesis aims to provide researchers and industry stakeholders with a clear roadmap for advancing the field and accelerating the deployment of intelligent, scalable, and sustainable robotic solutions for the future of agriculture.
As the crop with the longest planting history, widest planting area, and highest yield in hilly and mountainous areas, the improvement of the comprehensive mechanization level of rice cultivation and harvesting is of great significance for promoting the comprehensive and high-quality development of agricultural mechanization throughout the entire hilly and mountainous area. This study takes the key technology of rice mechanization in hilly and mountainous areas as the incision, and based on the characteristics of hilly and mountainous areas in China and the current development status of rice mechanization in hilly and mountainous areas, conducts a technological frontier and trend analysis around the production mechanization of rice tillage, planting, harvesting and other main links, analyzes the gap in research and development of rice mechanization technology and equipment at home and abroad, explores the challenges faced by the development of rice mechanization in hilly and mountainous areas in China under different links, and looks forward to the future development trend of rice mechanization in hilly and mountainous areas, in order to provide new ideas and directions for promoting the overall mechanization level of rice production, filling the gaps in the development of agricultural mechanization in hilly and mountainous areas, and building up China's strength in agriculture. This study proposes that, the comprehensive mechanization rate of rice cultivation and harvesting in hilly and mountainous areas of China in 2022 is 80.8%, and the development is unbalanced and insufficient among different links and regions; Compared with advanced countries, the research and application of rice cultivation and harvesting machinery in China started relatively late, with a weak foundation and low comprehensive performance of products. Especially, there is a relative shortage of lightweight, efficient, and highly stable rice cultivation and harvesting machinery for hilly and mountainous areas. Key technologies such as engine high-power low emission design and manufacturing technology, tool wear resistance and consumption reduction optimization design technology, and frame lightweight design technology still need to be further overcome; Lightweight, intelligent, and green technologies are the future directions and trends for the development of rice production equipment in hilly and mountainous areas of China.
Open Access
Research Article
Issue
To address the dynamic and uncertain challenges posed by climate variability, soil heterogeneity, water distribution, and other key factors in crop field production, intelligent decision support systems (IDSS) that integrate domain knowledge and multi-source data for irrigation, fertilization, pest and disease control, and field dynamic management, are of great importance in meeting modern agriculture’s demands for high precision, efficiency, and sustainability. By encompassing the development stages, typical practices, and technological pathways in China and developed countries, this paper summarizes the representative progress in Internet of Things (IoT), multimodal fusion, knowledge representation, reinforcement learning, reasoning, and practical applications in IDSS. Prominent research challenges include the lack of real-time or near-real-time sensor data, static domain knowledge, poor multimodal decision-making capability, weak cross-field generalization, and various implementation barriers, such as a vague definition of data governance, high costs of service infrastructure, and low user acceptance intention. To overcome these challenges, future research should prioritize the development of scalable, dynamic, robust, interpretable, and trustworthy multimodal IDSS, promote the formulation of standards, and establish an open platform for seamless model deployment, thereby facilitating the transformation from experience-driven to intelligence-driven agricultural production paradigms.
Citrus surface defects play a pivotal role in the fruit inspection and grading during agricultural production. Surface imperfections are also much easier to spot than inside ones, leading to accelerate the deterioration. However, conventional detection of citrus surface defects cannot fully meet the overall quality assessment in the large-scale production in recent years, due to the low efficiency and accuracy. In this study, an accurate, rapid and real-time detection was proposed to consider the diverse and complex nature of surface imperfections observed in citrus fruits. This speed and precision of detection were also enhanced for the quality of surface defects. Firstly, the images of citrus fruits were captured by industrial cameras. The generated images were enhanced to make the region of interest more outstanding. Then, the YOLOv7-CACT model was improved for the defect region in the enhanced citrus image. The coordinate attention (CA) module was introduced in the backbone network, in order to increase the attention to the defective part. The contextual transformer (CT) module was introduced in the head of the network to fuse the static and dynamic contextual representation features, thus enhancing the feature expression of the defective part. The superior performance was achieved in the modified YOLOv7-CACT model, compared with the baseline version. Especially, the detection accuracy was improved by 4.1% in the mean average precision (mAP). Consequently, the modified model was fully met the accuracy requirements for the identification of citrus surface defects in an offline setting. TensorRT was also employed around YOLOv7-CACT for the deployment of improved model, in order to real-time detect in practical scenarios. The results show that the improved YOLOv7-CACT-RT model was performed the best to detect the surface defects on the surface of citrus fruits in the grading and sorting production line with a transition rate of 10 fruits per second. The deployed YOLOv7-CACT-RT model was loaded into the grading software programmed by C++ language, in order to validate the performance. An online detection was conducted on 198 mixed normal and defective citrus fruits on the sorting line, achieving a detection accuracy of 94.4%. The improved model can be directly applied to grade and sort fruit in the production line, according to the external qualities. Meanwhile, this model can also be extended to the real-time surface defects detection of other fruits without specialized knowledge. Our future research will focus on the registration and fusion of RGB and NIR image, in order to improve the detection accuracy of fruit defects.
In the era of digital agriculture, real-time monitoring and predictive modeling of crop growth are paramount, especially in autonomous farming systems. Traditional crop growth models, often constrained by their reliance on static, rule-based methods, fail to capture the dynamic and multifactorial nature of vegetable crop growth. This research tried to address these challenges by leveraging the advanced reasoning capabilities of pre-trained large language models (LLMs) to simulate and predict vegetable crop growth with accuracy and reliability. Modeling the growth of vegetable crops within these platforms has historically been hindered by the complex interactions among biotic and abiotic factors.
The methodology was structured in several distinct phases. Initially, a comprehensive dataset was curated to include extensive information on vegetable crop growth cycles, environmental conditions, and management practices. This dataset incorporates continuous data streams such as soil moisture, nutrient levels, climate variables, pest occurrence, and historical growth records. By combining these data sources, the study ensured that the model was well-equipped to understand and infer the complex interdependencies inherent in crop growth processes. Then, advanced techniques was emploied for pre-training and fine-tuning LLMs to adapt them to the domain-specific requirements of vegetable crop modeling. A staged intelligent agent ensemble was designed to work within the digital twin platform, consisting of a central managerial agent and multiple stage-specific agents. The managerial agent was responsible for identifying transitions between distinct growth stages of the crops, while the stage-specific agents were tailored to handle the unique characteristics of each growth phase. This modular architecture enhanced the model's adaptability and precision, ensuring that each phase of growth received specialized attention and analysis.
The experimental validation of this method was conducted in a controlled agricultural setting at the Xiaotangshan Modern Agricultural Demonstration Park in Beijing. Cabbage (Zhonggan 21) was selected as the test crop due to its significance in agricultural production and the availability of comprehensive historical growth data. Over five years, the dataset collected included 4300 detailed records, documenting parameters such as plant height, leaf count, soil conditions, irrigation schedules, fertilization practices, and pest management interventions. This dataset was used to train the LLM-based system and evaluate its performance using ten-fold cross-validation. The results of the experiments demonstrating the efficacy of the proposed system in addressing the complexities of vegetable crop growth modeling. The LLM-based model achieved 98% accuracy in predicting crop growth degrees and a 99.7% accuracy in identifying growth stages. These metrics significantly outperform traditional machine learning approaches, including long short-term memory (LSTM), XGBoost, and LightGBM models. The superior performance of the LLM-based system highlights its ability to reason over heterogeneous data inputs and make precise predictions, setting a new benchmark for crop modeling technologies. Beyond accuracy, the LLM-powered system also excels in its ability to simulate growth trajectories over extended periods, enabling farmers and agricultural managers to anticipate potential challenges and make proactive decisions. For example, by integrating real-time sensor data with historical patterns, the system can predict how changes in irrigation or fertilization practices will impact crop health and yield. This predictive capability is invaluable for optimizing resource allocation and mitigating risks associated with climate variability and pest outbreaks.
The study emphasizes the importance of high-quality data in achieving reliable and generalizable models. The comprehensive dataset used in this research not only captures the nuances of cabbage growth but also provides a blueprint for extending the model to other crops. In conclusion, this research demonstrates the transformative potential of combining large language models with digital twin technology for vegetable crop growth modeling. By addressing the limitations of traditional modeling approaches and harnessing the advanced reasoning capabilities of LLMs, the proposed system sets a new standard for precision agriculture. Several avenues also are proposed for future work, including expanding the dataset, refining the model architecture, and developing multi-crop and multi-region capabilities.
In modern agriculture, the rapid and accurate detection of chillies at different maturity stages is a critical step for determining the optimal harvesting time and achieving intelligent sorting of field-grown chillies. However, existing target detection models face challenges in efficiency and accuracy when applied to the task of detecting chilli maturity, which limit their widespread use and effectiveness in practical applications. To address these challenges, a new algorithm, Chilli-YOLO, was proposed for achieving efficient and precise detection of chilli maturity in complex environments.
A comprehensive image dataset was collected, capturing chillis under diverse and realistic agricultural conditions, including varying lighting conditions, camera angles, and background complexities. These images were then meticulously categorized into four distinct maturity stages: Immature, transitional, mature, and dried. Data augmentation techniques were employed to expand the dataset and enhance the model's generalization capabilities. To develop an accurate and efficient chili maturity detection system, the YOLOv10s object detection network was chosen as the foundational architecture. The model's performance was further enhanced through strategic optimizations targeting the backbone network. Specifically, standard convolutional layers were replaced with Ghost convolutions. This technique generated more feature maps from fewer parameters, resulting in significant computational savings and improved processing speed without compromising feature extraction quality. Additionally, the C2f module was substituted with the more computationally efficient GhostConv module, further reducing redundancy and enhancing the model's overall efficiency. To improve the model's ability to discern subtle visual cues indicative of maturity, particularly in challenging scenarios involving occlusion, uneven lighting, or complex backgrounds, the partial self-attention (PSA) module within YOLOv10s was replaced with the second-order channel attention (SOCA) mechanism. SOCA leverages higher-order feature correlations to more effectively capture fine-grained characteristics of the chillis. This enabled the model to focus on relevant feature channels and effectively identify subtle maturity-related features, even when faced with significant visual noise and interference. Finally, to refine the precision of target localization and minimize bounding box errors, the extended intersection over union (XIoU) loss function was integrated into the model training process. XIoU enhances the traditional IoU loss by considering factors such as the aspect ratio difference and the normalized distance between the predicted and ground truth bounding boxes. By optimizing for these factors, the model achieved significantly improved localization accuracy, resulting in a more precise delineation of chillis in the images and contributing to the overall enhancement of the detection performance. The combined implementation of these improvements aimed to construct an effective approach to correctly classify the maturity level of chillis within the challenging and complex environment of a real-world farm.
The experimental results on the custom-built chilli maturity detection dataset showed that the Chilli-YOLO model performed excellently across multiple evaluation metrics. The model achieved an accuracy of 90.7%, a recall rate of 82.4%, and a mean average precision (mAP) of 88.9%. Additionally, the model's computational load, parameter count, model size, and inference time were 18.3 GFLOPs, 6.37 M, 12.6 M, and 7.3 ms, respectively. Compared to the baseline model, Chilli-YOLO improved accuracy by 2.6 percent point, recall by 2.8 percent point and mAP by 2.8 percent point. At the same time, the model's computational load decreased by 6.2 GFLOPs, the parameter count decreased by 1.67 M, model size reduced by 3.9 M. These results indicated that Chilli-YOLO strikes a good balance between accuracy and efficiency, making it capable of fast and precise detection of chilli maturity in complex agricultural environments. Moreover, compared to earlier versions of the YOLO model, Chilli-YOLO showed improvements in accuracy of 2.7, 4.8, and 5 percent point over YOLOv5s, YOLOv8n, and YOLOv9s, respectively. Recall rates were higher by 1.1, 0.3, and 2.3 percent point, and mAP increased by 1.2, 1.7, and 2.3 percent point, respectively. In terms of parameter count, model size, and inference time, Chilli-YOLO outperformed YOLOv5. This avoided the issue of YOLOv8n's lower accuracy, which was unable to meet the precise detection needs of complex outdoor environments. When compared to the traditional two-stage network Faster RCNN, Chilli-YOLO showed significant improvements across all evaluation metrics. Additionally, compared to the one-stage network SSD, Chilli-YOLO achieved substantial gains in accuracy, recall, and mAP, with increases of 16.6%, 12.1%, and 16.8%, respectively. Chilli-YOLO also demonstrated remarkable improvements in memory usage, model size, and inference time. These results highlighted the superior overall performance of the Chilli-YOLO model in terms of both memory consumption and detection accuracy, confirming its advantages for chilli maturity detection.
The proposed Chilli-YOLO model optimizes the network structure and loss functions, not only can significantly improve detection accuracy but also effectively reduce computational overhead, making it better suites for resource-constrained agricultural production environments. The research provides a reliable technical reference for intelligent harvesting of chillies in agricultural production environments, especially in resource-constrained settings.
As agriculture increasingly relies on technological innovations to boost productivity and ensure sustainability, farmers need efficient and accurate tools to aid their decision-making processes. A key challenge in this context is the retrieval of specialized agricultural knowledge, which can be complex and diverse in nature. Traditional agricultural knowledge retrieval systems have often been limited by the modalities they utilize (e.g., text or images alone), which restricts their effectiveness in addressing the wide range of queries farmers face. To address this challenge, a specialized multimodal question-answering system tailored for cabbage cultivation was proposed. The system, named Agri-QA Net, integrates multimodal data to enhance the accuracy and applicability of agricultural knowledge retrieval. By incorporating diverse data modalities, Agri-QA Net aims to provide a holistic approach to agricultural knowledge retrieval, enabling farmers to interact with the system using multiple types of input, ranging from spoken queries to images of crop conditions. By doing so, it helps address the complexity of real-world agricultural environments and improves the accessibility of relevant information.
The architecture of Agri-QA Net was built upon the integration of multiple data modalities, including textual, auditory, and visual data. This multifaceted approach enables the system to develop a comprehensive understanding of agricultural knowledge, allowed the system to learn from a wide array of sources, enhancing its robustness and generalizability. The system incorporated stateof-the-art deep learning models, each designed to handle one specific type of data. Bidirectional Encoder Representations from Transformers (BERT)'s bidirectional attention mechanism allowed the model to understand the context of each word in a given sentence, significantly improving its ability to comprehend complex agricultural terminology and specialized concepts. The system also incorporated acoustic models for processing audio inputs. These models analyzed the spoken queries from farmers, allowing the system to understand natural language inputs even in noisy, non-ideal environments, which was a common challenge in real-world agricultural settings. Additionally, convolutional neural networks (CNNs) were employed to process images from various stages of cabbage growth. CNNs were highly effective in capturing spatial hierarchies in images, making them well-suited for tasks such as identifying pests, diseases, or growth abnormalities in cabbage crops. These features were subsequently fused in a Transformer-based fusion layer, which served as the core of the Agri-QA Net architecture. The fusion process ensured that each modality—text, audio, and image—contributes effectively to the final model's understanding of a given query. This allowed the system to provide more nuanced answers to complex agricultural questions, such as identifying specific crop diseases or determining the optimal irrigation schedules for cabbage crops. In addition to the fusion layer, cross-modal attention mechanisms and domain-adaptive techniques were incorporated to refine the model's ability to understand and apply specialized agricultural knowledge. The cross-modal attention mechanism facilitated dynamic interactions between the text, audio, and image data, ensuring that the model paid attention to the most relevant features from each modality. Domain-adaptive techniques further enhanced the system's performance by tailoring it to specific agricultural contexts, such as cabbage farming, pest control, or irrigation management.
The experimental evaluations demonstrated that Agri-QA Net outperforms traditional single-modal or simple multimodal models in agricultural knowledge tasks. With the support of multimodal inputs, the system achieved an accuracy rate of 89.5%, a precision rate of 87.9%, a recall rate of 91.3%, and an F1-Score of 89.6%, all of which are significantly higher than those of single-modality models. The integration of multimodal data significantly enhanced the system's capacity to understand complex agricultural queries, providing more precise and context-aware answers. The addition of cross-modal attention mechanisms enabled for more nuanced and dynamic interaction between the text, audio, and image data, which in turn improved the model's understanding of ambiguous or context-dependent queries, such as disease diagnosis or crop management. Furthermore, the domain-adaptive technique enabled the system to focus on specific agricultural terminology and concepts, thereby enhancing its performance in specialized tasks like cabbage cultivation and pest control. The case studies presented further validated the system's ability to assist farmers by providing actionable, domain-specific answers to questions, demonstrating its practical application in real-world agricultural scenarios.
The proposed Agri-QA Net framework is an effective solution for addressing agricultural knowledge questions, especially in the domain of cabbage cultivation. By integrating multimodal data and leveraging advanced deep learning techniques, the system demonstrates a high level of accuracy and adaptability. This study not only highlights the potential of multimodal fusion in agriculture but also paves the way for future developments in intelligent systems designed to support precision farming. Further work will focus on enhancing the model's performance by expanding the dataset to include more diverse agricultural scenarios, refining the handling of dialectical variations in audio inputs, and improving the system's ability to detect rare crop diseases. The ultimate goal is to contribute to the modernization of agricultural practices, offering farmers more reliable and effective tools to solve the challenges in crop management.
Open Access
Issue
In order to address the challenge of non-destructive detection of tomato fruit ripeness in controlled environments, this study proposed a real-time instance segmentation method based on the edge device. This method combined the principles of phenotype robots and machine vision based on deep learning. A compact and remotely controllable phenotype detection robot was employed to acquire precise data on tomato ripeness. The video data were then processed by using an efficient backbone and the FeatFlowNet structure for feature extraction and analysis of key-frame to non-key-frame mapping from video data. To enhance the diversity of training datasets and the generalization of the model, an innovative approach was chosen by using random enhancement techniques. Besides, the PolyLoss optimization technique was applied to further improve the accuracy of the ripeness multi-class detection tasks. Through validation, the method of this study achieved real-time processing speeds of 90.1 fps (RTX 3070Ti) and 65.5 fps (RTX 2060 S), with an average detection accuracy of 97% compared to manually measured results. This is more accurate and efficient than other instance segmentation models according to actual testing in a greenhouse. Therefore, the results of this research can be deployed in edge devices and provide technical support for unmanned greenhouse monitoring devices or fruit-picking robots in facility environments.
京公网安备11010802044758号